Heteroscedasticity and Autocorrelation Carlo Favero Favero () Heteroscedasticity and Autocorrelation 1 / 17
Heteroscedasticity, Autocorrelation, and the GLS estimator Let us reconsider the single equation model and generalize it to the case in which the hypotheses of diagonality and constancy of the conditional variances-covariance matrix of the residuals do not hold: y = Xβ + ɛ, (1) ɛ n.d. 0, σ 2 Ω, where Ω is a (T T) symmetric and positive definite matrix. When the OLS method is applied to model (1), it delivers estimators which are consistent but not efficient; moreover, the traditional formula for the variance-covariance matrix of the OLS estimators, σ 2 (X 0 X) 1, is wrong and leads to an incorrect inference. Favero () Heteroscedasticity and Autocorrelation 2 / 17
Heteroscedasticity, Autocorrelation, and the GLS estimator Using the standard algebra, it can be shown that the correct formula for the variance-covariance matrix of the OLS estimator is: σ 2 X 0 X 1 X 0 ΩX X 0 X 1. To find a general solution to this problem, remember that the inverse of a symmetric definite positive matrix is also symmetric and definite positive and that for a given matrix Φ, symmetric and definite positive, there always exists a (T T) non-singular matrix K, such that K 0 K =Ω 1 and KΩK 0 = I T. Favero () Heteroscedasticity and Autocorrelation 3 / 17
Heteroscedasticity, Autocorrelation, and the GLS estimator Consider the regression model obtained by pre-multiplying both the right-hand and the left-hand sides of (1) by K: Ky = KXβ + Kɛ, (2) Kɛ n.d. 0, σ 2 I T. The OLS estimator of the parameters of the transformed model (2) satisfies all the conditions for the applications of the Gauss Markov theorem; therefore, the estimator bβ GLS = X 0 K 0 KX 1 X 0 K 0 Ky 1 = X 0 Ω X 1 X 0 Ω 1 y, known as the generalised least squares (GLS) estimator, is BLUE. The variance of the GLS estimator, conditional upon X, becomes 1 Var b β GLS j X = Σ β = σ 2 X 0 Ω X 1. Favero () Heteroscedasticity and Autocorrelation 4 / 17
Heteroscedasticity, Autocorrelation, and the GLS estimator Consider, for example, the variances of the OLS and the GLS estimators. Using the fact that if A and B are positive definite and A B is positive semidefinite, then B 1 A 1 is also positive semidefinite, we have: X 0 Ω 1 X X 0 X X 0 ΩX 1 X 0 X = X 0 K 0 KX X 0 X X 0 K 1 K 0 1 1 X X 0 X = X 0 K 0 I K 0 1 X X 0 K 1 K 0 1 1 0 X X K 1 KX where = X 0 K 0 M 0 W M WKX, M W = I W W 0 W 1 W 0, (3) W = K 0 1 X. (4) Favero () Heteroscedasticity and Autocorrelation 5 / 17
Heteroscedasticity, Autocorrelation, and the GLS estimator The applicability of the GLS estimator requires an empirical specification for the matrix K. We consider here three specific applications where the appropriate choice of such a matrix leads to the solution of the problems in the OLS estimator generated, respectively, by the presence of first-order serial correlation in the residuals, by the presence of heteroscedasticity in the residuals by the presence of both of them. Favero () Heteroscedasticity and Autocorrelation 6 / 17
Correction for Serial Correlation (Cochrane-Orcutt) Consider first the case of first-order serial correlation in the residuals. We have the following model: y t = x 0 tβ+u t, u t = ρu t 1 + ɛ t, ɛ t n.i.d. 0, σ 2 ɛ, which, using our general notation, can be re-written as: y = Xβ + ɛ, (5) ɛ n.d. 0, σ 2 Ω, σ 2 = σ 2 ɛ 1 ρ 2, (6) Favero () Heteroscedasticity and Autocorrelation 7 / 17
Correction for Serial Correlation (Cochrane-Orcutt) 2 = 6 4 1 ρ ρ 2.. ρ T 1 ρ 1 ρ.. ρ T 2 ρ 2. 1......... ρ T 2.. ρ 1 ρ ρ T 1 ρ T 2.. ρ 1 In this case, the knowledge of the parameter ρ allows the empirical implementation of the GLS estimator. An intuitive procedure to implement the GLS estimator can then be the following: 1 estimate the vector β by OLS and save the vector of residuals bu t ; 2 regress bu t on bu t 1 to obtain an estimate bρ of ρ; 3 construct the transformed model and regress (y t bρy t 1 ) on (x t bρx t 1 ) to obtain the GLS estimator of the vector of parameters of interest. Note that the above procedure, known as the Cochrane Orcutt procedure, can be repeated until convergence. Favero () Heteroscedasticity and Autocorrelation 8 / 17 3. 7 5
Correction for Heteroscedasticity (White) In the case of heteroscedasticity, our general model becomes y = Xβ + ɛ, ɛ n.d. (0,Ω), 2 σ 2 3 1 0 0.. 0 0 σ 2 2 0.. 0 Ω =...... 6....... 7 4 0.. 0 σ 2 T 1 0 5 0 0.. 0 σ 2 T In this case, to construct the GLS estimator, we need to model heteroscedasticity choosing appropriately the K matrix. Favero () Heteroscedasticity and Autocorrelation 9 / 17
Correction for Heteroscedasticity (White) White (1980) proposes a specification based on the consideration that in the case of heteroscedasticity the variance-covariance matrix of the OLS estimator takes the form: σ 2 X 0 X 1 X 0 ΩX X 0 X 1, which can be used for inference, once an estimator for Φ is available. The following unbiased estimator of Φ is proposed: 2 bω= 6 4 bu 2 1 0 0.. 0 0 bu 2 2 0.. 0............ 0.. 0 bu 2 T 1 0 0 0.. 0 bu 2 T 3. 7 5 Favero () Heteroscedasticity and Autocorrelation 10 / 17
Correction for Heteroscedasticity (White) This choice for bω leads to the following degrees of freedom corrected heteroscedasticity consistent parameters covariance matrix estimator: Σ W β = T T k X0 X! T 1 bu 2 t X t X 0 t X 0 X 1 t=1 This estimator corrects for the OLS for the presence of heteroscedasticity in the residuals without modelling it. Favero () Heteroscedasticity and Autocorrelation 11 / 17
Correction for heteroscedasticity and serial correlation (Newey-West) The White covariance matrix assumes serially uncorrelated residuals. Newey and West(1987) have proposed a more general covariance estimator that is robust to heteroscedasticity and autocorrelation of the residuals of unknow form. This HAC (heteroscedasticity and autocorrelation consistent) coefficient covariance estimators is given by: Σ NW β = X 0 X 1 T ˆ Ω X 0 X 1 where ˆ Ω is a long-run covariance estimators Favero () Heteroscedasticity and Autocorrelation 12 / 17
Correction for heteroscedasticity and serial correlation (Newey-West) bω = bγ (0) + bγ (j) = T t=1 p j=1 1 bu t bu t j X t X 0 t j j hb Γ (j) + bγ ( j)i, (7) p + 1! 1 T note that in absence of serial correlation bω = bγ (0) and we are back to the White Estimator. Implementation of the estimator requires a choice of p, which is the maximum lag at whcih correlation is still present. The weighting scheme adopted guarantees a positive definite estimated covariance matrix by multiplying the sequence of the bγ (j) s by a sequence of weights that decreases as jjj increases. Favero () Heteroscedasticity and Autocorrelation 13 / 17
Fama-MacBeth(1973) Consider the 25 portfolios and run for each of them the CAPM regression over the sample 1962:1 2014:6. These regressions deliver 25 betas. Take now a second-step regression in which the cross-section of the average (over the sample 1962:1-2014:6) monthly returns on the 25 portfolios are projected on the 25 betas: r i = γ 0 + γ 1 β i + u i Under the null of the CAPM i) residuals should be randomly distributed around the regression line, ii) γ 0 = E r f, γ 1 = E r m r f Favero () Heteroscedasticity and Autocorrelation 14 / 17
Fama-MacBeth(1973) Favero () Heteroscedasticity and Autocorrelation 15 / 17
Fama-MacBeth(1973) The cross-sectional regression strongly rejects the CAPM. Note however that this regression is affected by an inference problem caused by the correlation of residuals in the cross-section regression. Fama and MacBeth (1973) address this problem by estimating month-by-month cross-section regressions of monthly returns on the betas obtained on the full sample. The time series means of the monthly slopes and intercepts, along with the standard errors of the means, are then used to run the appropriate tests Favero () Heteroscedasticity and Autocorrelation 16 / 17
Fama-MacBeth(1973) The application of the Fama-MacBeth on the sample 1962:1 2014:6 delivers the following results TABLE 3.5: Statistics on the distribution of coefficients from Fama-MacBeth γ 0 γ 1 Mean 2.07-0.807 St.Dev 9.56 10.06 Obs 630 630 t-stat 5.437-2.01 An alternative route would be to construct Heteroscedasticity and Autocorrelation Consistent(HAC) estimators. Favero () Heteroscedasticity and Autocorrelation 17 / 17